1. Light-Cone Averaging and Dispersion of the d L – z Relation Ido Ben-Dayan DESY 1202.1247, 1207.1286, 1209.4326, 1302.0740 + Work in Progress IBD, M. Gasperini, G. Marozzi, F. Nugier, G. Veneziano April 3 rd 2013, Tel-Aviv University

2. Outline Background Light-Cone Averaging Prescription Application: Luminosity-distance (d L ) Redshift (z) Relation in the Concordance Model. Lensing Dispersion of d L Probing the Primordial Power Spectrum

3. Averaging in Physics Electromagnetism: Maxwell’s equations are linear. Averaging and solving commute. Microscopic E&M is smoothly averaged to Macroscopic E&M. GR is non-linear. Averaging and solving do not commute.

4. Main stream Cosmology Assume FLRW = homogeneous and isotropic metric.  Implicit averaging Modeling the energy momentum tensor as a perfect fluid. Pert. give rise to structure, highly non-linear at some scale. Background unchanged.  Implicitly neglected the possibility of backreaction.

6. Consistency? SDSS 2.73 K PLANCK

7. Perturbations at Low Redshift Measurements of SNIa  Mostly neglected, naively argued as irrelevant ~10 -10 The concordance model of cosmology: (before PLANCK) ~73% of the critical energy density is not accounted for by known matter, dark matter or curvature. PLANCK: DE is only 68% of the critical energy density!

8. Universe Composition Today CC/DE becomes relevant only at z~1, Coincidence Problem? Based on CMB, LSS and SNIa observations.

9. 2011 Nobel Prize in Physics The 2011 Nobel Prize in Physics is awarded "for the discovery of the accelerating expansion of the Universe through observations of distant supernovae" with one half to Saul Perlmutter and the other half jointly to Brian P. Schmidt and Adam G. Riess2011 Nobel Prize in PhysicsSaul PerlmutterBrian P. SchmidtAdam G. Riess

10. Average d L on the past LC as a function of redshift A= redshift, V= light-cone coordinate

11. LC Averaging Useful for interpreting light-like signals in cosmological observations. Hyper-surfaces using meaningful physical quantities: Redshift, temperature etc. Observations are made on the light- cone. Volume averaging give artefacts and the matching with data is not clear. Past attempts: Coley 0905.2442; Rasanen 1107.1176, 0912.3370

12. Light Cone Averaging 1104.1167 A-priori - the averaging is a geometric procedure, does not assume a specific energy momentum tensor. The prescription is gauge inv., {field reparam. A->A’(A), V->V’(V)} and invariant under general coordinate transformation. A(x) is a time-like scalar, V(x) is null. This gives a procedure for general space-times.

13. GLC Metric and Averages Ideal Observational Cosmology – Ellis et al. Evaluating scalars at a constant redshift for a geodetic observer.

14. GLC Metric FLRW τ can be identified as the time coordinate in the synchronous gauge of arbitrary space-time. IBD et al. ’12

15. (z), (z) A= redshift, V= light-cone coordinate

16. Averaged d L at Constant Redshift Novelty: In principle, exact treatment of the geodesic equations and the averaging hyper-surface for any spacetime and any DE model, as long as the geodesic equation is unchanged. Previous attempts are limited to perturbations about FLRW and had to solve order by order: Vanderveld et al. – post Newtonian, Barausse et al. 2005, Kolb et al. 2006 – SG superhorizon, Pyne at al 2005… Rebuttal : Hirata et al., Geshnizjani et al. We work up to 2 nd order in perturbation theory in the Poisson Gauge.

17. Averaged d L at Constant Redshift Ethrington’s Reciprocity Law, for any spacetime:

18. Functions of d L Overbars and {…} denote ensemble average, denote LC average. Averages of different functions of scalars receive different contributions.

19. The Perturbed Quantities We work up to 2 nd order in perturbation theory in the Poisson Gauge. Both the area distance and the measure of integration are expressed in terms of the gravitational potential and its derivatives. Vector and tensor pert. do not contribute.

20. Statistical Properties In principle, we can now calculate (z) to first order in the gravitational potential ~ void model. In order not to resort to a specific realization we need LC+statistical/ensemble average. If perturbations come from primordial Gaussian fluc. (inflation)

21. LC average of flux for any space-time amounts to the area of the 2-sphere! The Optimal Observable - Flux

22. Linear PS of the grav. potential

23. Interpretation & Analysis d L is a stochastic observable – mean, dispersion, skewness... In the flux - The dominant contribution are Doppler terms ~k 2 Any other function of d L gets also k 3 contributions – lensing contribution, dominates at large redshift, z>0.3 In principle the upper limit can be infinite. In practice, until where do we trust our spectrum? Linear treatment k<0.1-1 Mpc -1 and non-linear treatment k<30h Mpc -1

24. “Doppler 2 ” term in Flux (CDM) ∞ 1 Mpc -1 0.1 Mpc -1

25. ↵ Lensing 2 Term – Not in Flux (CDM) ∞ 1 Mpc -1 0.1 Mpc -1

26. Interpretation & Analysis Superhorizon scales are subdominant. At small enough scales, the transfer function wins. At intermediate scales, the phase space factor competes with the initial small amplitude. In the non-linear regime we use a fit from simulations. (Takahashi et al. 1208.2701)

27. Fractional corrections to the Flux and d L, k UV =30h Mpc -1

28. Distance Modulus Average and Dispersion, at z>0.5, Lensing Dominates!

29. Lensing Dispersion

30. Closer Look on the Lensing Integral

31. Lensing Dispersion Current Data up to z~1.3 Most Conservative Approach: Constrains late time power spectrum and/or numerical simulations. h(k,z) incorporates the dependence on the EOS parameter w(z) etc. The dispersion can be used to constrain EOS, primordial power spectrum etc.

32. Primordial Power Spectrum

33. Constraints from Spectral Distortions Chluba, Erickcek, IBD 2012:

34. Lensing Dispersion Current data has SN up to z~1.3 Statistical Analysis: dispersion<0.12 in mag. (March et al. 2011) Extended halofit, Inoue & Takahashi 2012, up to k=320h Mpc -1  Can constrain the amplitude on scales 1<k<30h to roughly 10 -7 and better.

35. Results and Lessons Unlike volume averages: No divergences The contribution from inhomogeneities is several orders of magnitude larger than the naïve expectations due to the large phase space factor. The size of the contribution (f d, f Φ ) is strongly dependent on the quantity whose average is considered. Our approach is useful whenever dealing with information carried by light-like signals travelling along our past light-cone.

36. Conclusions Flux is the optimal observable. Different bias or “subtraction” mechanisms, in order to extract cosmological parameters. Irreducible Scatter - The dispersion is large ~ 2-10% Λ CDM, of the critical density depending on the spectrum. Scatter is mostly from the LC average. It gives theoretical explanation to the intrinsic scatter of SN measurements. The effect is too SMALL AND has the WRONG z dependence to simulate observable CC!

37. Open Issues/Future Prospects 1. Using the lensing dispersion to constrain the power spectrum. 2. Matching the effect to other probes: CMB, LSS 3. Applying LC averaging to cosmic shear, BAO, kSZ, strong lensing etc. 4. Other applications, averaging of EFE ….Many open theoretical and pheno. problems.

38. k max =0.1 Mpc -1, CDM

39. k max =1 Mpc -1

40. Prescription Properties Dynamical properties: Generalization of Buchert-Ehlers commutation rule: For actual physical calculations, use EFE/ energy momentum tensor for evaluation. Example: which gravitational potential to use in evaluating the d L -z relation. Averages of different functions give different outcome

41. Summary & Conclusions Application of light cone averaging formalism to the d L -z relation. INHOMOGENEITIES CANNOT FAKE DARK ENERGY AT THE OBSERVABLE LEVEL! The effect of averaging can, in principle, be distinguished from the homogeneous contribution of CC/DE. Variance gives theoretical explanation to the intrinsic scatter of SN measurements.

42. Open Issues/Future Prospects 1. Estimates of the non-linear regime, especially 0.1 Mpc -1 <k< 1 Mpc -1. 2. Cosmological parameter analysis. 3. Matching the effect to other probes: CMB, LSS 4. Applying LC averaging to BAO, kSZ and many more. 5. Other applications, averaging of EFE ….Many open theoretical and pheno. problems.

43. GLC to FLRW NG 1 st Order

44. LC Calculation and LCDM Pure FLRW: Transform from the GLC metric to the longitudinal gauge Perturbed:

45. Measure of Integration

46. Statistical Properties In principle, we can now calculate (z) to first order in the gravitational potential ~ void model. In order not to resort to a specific realization we need LC+statistical/ensemble average. If perturbations come from primordial Gaussian fluc. (inflation)

47. BR of Statistical and LC Averaging The mean of a scalar: => Effects are second order, but we have the full backreaction of the inhomogeneities of the metric at this order! The variance to leading order:

48. Dominant Terms Doppler effect due to the perturbation of the geodesic. The Lensing Term:

49. Conclusions GR + Standard Pert. Theory + Averaging challenge the concordance model. Perturbations cannot be discarded as negligible. Any explanation of the cosmic acceleration will have to take them into account.

51. CC/ Dark Energy/ Modified Gravity/ Voids

52. GLC to FLRW NG 1 st Order

53. LC Calculation and LCDM Pure FLRW: Perturbed: Comparing by defining an effective redshift and averaging at constant redshift and w=w 0

54. Power Spectrum

55. “Doppler 2 ” term ∞ 1 Mpc -1 0.1 Mpc -1

56. Lensing 2 Term ↵ ∞ 1 Mpc -1 0.1 Mpc -1

57. Non-Trivial Averages Off-Center LTB Anisotropic Models (Except Kantowski-Sachs) More general metrics. Perturbed FLRW Application: calculating the averaged luminosity – distance redshift relation Past attempts: Vanderveld et al. – post Newtonian, Barausse et al., Kolb et al. – SG superhorizon, Pyne at al.,